توسعه نمودارهاي كنترل بهبوديافته مبتني بر تكنيك نمونه‌گيري تكرار‌شونده براي پايش جريان داده‌هاي با ابعاد بالا با در نظر گرفتن اثر مشاهدات آلوده به خطاي اندازه‌گيري

In todayʹs highly competitive industrial environment, gaining market share an‎d improving product quality increasingly necessitate monitoring a large number of quality characteristics. On the other han‎d, despite recent advancements in data acquisition technologies, measurement error remains unavoidable in practice, thereby increasing uncertainty in the collected quality data. This study addresses two primary objectives: first, to accelerate the detection of covariance matrix shifts in high-dimensional processes, an‎d second, to account for the effect of imprecise observations on chart performance. To achieve this, the adaptive thresholding Lasso (ATL) chart is first equipped with a repetitive sampling scheme. Subsequently, an updat‎ed version of the proposed chart is developed using an additive covariate model to account for the effects of measurement error. The performance of the proposed chart is compared to the classical adaptive thresholding Lasso chart through Monte Carlo simulations. This eva‎luation uses average run length (ARL), stan‎dard deviation of run length (SDRL), an‎d Average Sample Number (ASN) metrics across seven out-of-control scenarios, two process dimensions, four shift magnitudes, an‎d five error variances. Simulation results indicate that the proposed chart consistently outperforms the classical ATL chart, achieving faster an‎d more stable detection of covariance matrix disturbances, as evidenced by average ARL reductions of up to 35% an‎d lower SDRL values across most settings investigated. The results further indicate that measurement system error degrades the performance of both the improved an‎d classical charts in response to covariance matrix disturbances; however, the proposed method maintains its detection advantage even with error-contaminated observations. Finally, the applicability of the proposed control charting method is validated using real dimensional inspection data from an aluminum intake manifold.

محمدرضا ملكي

مرتضي هادي

معصومه مسي بيدگلي , محمد كاظم زاده